GoldVarb
Version 2
A Variable Rule Application
for the Macintosh
David Rand
and
David
Sankoff
GoldVarb
Version
2.0
April,
1990
For free distribution only
To
request information or make comments contact:
D. Sankoff
or D. Rand
Centre
de recherches mathmatiques
Universit
de Montral
C.P.
6128, Succursale A, Montral H3C 3J7
(514)
343-7574 or -7271
E-mail:
RAND@CC.UMontreal.CA
RAND@UMTLVR.BITNET
Table of Contents
1. Introduction
2. Data and Results
3. Getting Started
4. Token Files
4.1 Creating a New Token
File
4.2 Importing a
Non-Native Token File
4.3 Creating a Token
File from an Existing Token File
4.4 Searching and
Replacing
5. Condition Files
5.1 Using the Recode setup
dialogue box
5.2 The Application
Values
6. Cell Files
7. Variable rule analysis
8. Using the Menus
8.1 The � Menu
8.2 The File Menu
8.3 The Edit Menu
8.4 The Tokens Menu
8.5 The Cells Menu
8.6 The Window Menu
8.7 The Font & Style
Menus
9. Glossary
Appendix
I. More on condition files
Appendix
II. The Macintosh character set
1. INTRODUCTION
GoldVarb is a Macintosh application for carrying out
variable rule analysis and associated data manipulations and displays. It is
based on programs previously circulated by David Sankoff, Pascale Rousseau, Don
Hindle and Susan Pintzuk, but, as well as containing many new features, it has
been completely restructured and reprogrammed in PASCAL by David Rand.
Successive versions have been extensively tested by researchers in the
linguistics departments of the University of Pennsylvania and the University of
Ottawa and in the Dpartement d'anthropologie of the Universit de Montral.
GoldVarb 1.6 was distributed in October of 1988 at the XVIIth NWAVE
colloquium held at the Centre de recherches mathmatiques of the Universit de
Montral.[1] This manual describes version 2.0 which includes
improvements implemented in the intervening year and a half, as well as several
extensions to the user interface.
This manual describes only how to use GoldVarb. The
underlying mathematics as well as the linguistic interpretation and pertinence
of variable rule analysis are discussed elsewhere.[2] GoldVarb is a stand-alone application, requiring no
other software other than the operating system. The text-editing capabilities
of GoldVarb were adapted from the text-editor Utile+[3] which interested users can obtain from its author.
We assume that the reader has some experience with at
least one other Macintosh application. The glossary in 9 contains a few tips
for beginners. It is important to use a recent version of the system software[4]: that is, the System, Finder and
MultiFinder files, and other
associated files.
GoldVarb may be run on any member of the Macintosh
family of computers. Nevertheless, it is not recommended for use on models
older than the Macintosh Plus.
2. DATA AND RESULTS
The data for variable rule analysis consist of a list
of tokens coded for a certain
number of factors. Figure 1 shows part of a GoldVarb data file.
Figure
1.
This type of file will be discussed in more detail in
4. The data in the figure bear on the phenomenon of plural morpheme expression
in Nepean noun phrases. Each line starting with an open parenthesis contains
the information on one adjective, noun or determiner token from a corpus of
informal interaction among Nepean adolescent gang members. The 1s and 0s
in the first column indicate the presence or absence of the plural marker –enmas. The a, n and d refer to the grammatical
category (adjective, noun or determiner) of the word eligible to be marked. The
c and s in the third column indicate whether the token comes from an object
or subject noun phrase, respectively.
Figure
2.
The key output of a variable rule analysis consists
of a list of numbers, one associated with each factor which may affect the
variable being studied. Figure 2 contains part of a GoldVarb output file (see
7 for more detail) from an analysis of the data on Nepean plural marker
expression of which the tokens in Figure 1 form part. The variable is plural marker presence or absence, the factors
are a, n, d, c and s,
grouped into two factor groups,
one containing a, n and d and the other c and s. The numbers are
factor weights, indicating in
this instance the degree to which the factor favours marker presence rather
than deletion. Higher numbers indicate that the corresponding factor favours
marker presence more than the other factor(s) within the same factor group. The
input is a kind of average
tendency for marker presence. The precise interpretation of all these numbers
can be found in the document Variable Rules2.
3. GETTING STARTED
The different icons associated with different types
of GoldVarb files are illustrated in Figure 3. The program comes with a sample token
file, entitled Nepean.Tok, which
we will use to illustrate its functioning. In this section we will show how to
get from input to output in the simplest cases. In succeeding sections we
discuss the various steps and options in more detail.
Figure
3.
Open the file Nepean.Tok by double-clicking on its
icon; this will automatically open GoldVarb. The screen should appear as in
Figure 4. Nepean.Tok contains data on Nepean plural expression; indeed Figure 1
portrays part of this same file. The figure in the lower left corner of its
window indicates the largest block of memory still available to GoldVarb[5]. Ignore for the moment the Factor specification box associated
with the token file. Token files and factor specifications will be discussed in
more detail in 4.
We will use the information contained in Nepean.Tok
as is. That is, we will not recode any of the factors. Thus we choose No
recode in the Tokens menu. A dialogue box will appear asking for a name
for a condition file. Type
Nepean.Cnd (or whatever name you wish) and click Save. The program will produce a default condition file
as in Figure 5. As with the token document, available memory is indicated in
the lower left corner of the window. The second line of this document is just a
comment, as indicated by the semi-colon in the first column. Note that the
token file remains open. Condition files will be discussed in more detail in
5.
Figure
4.
Figure
5.
The next step is to construct a cell file by choosing Load cells to memory in the Cells menu.[6] A series of dialogue boxes will appear to establish
options and to name files. The first simply requires confirmation that the
cells be constructed on the basis of the currently open token and condition
files. After clicking Yes, the
user is requested to name the cell file, for example Nepean.Cel. The tokens
in Nepean.Tok are counted as they are checked for consistency with the values
declared in the Factor specification box discussed in 4. A message appears when this is
complete. Click OK to continue. A
third dialogue box then appears, asking about application and non-application
values. Ignore this for the moment; it will be discussed in 5. Just click OK to accept the default values furnished by the
program. Finally, a fourth dialogue box asks for a name for a result file. After the user clicks Save, GoldVarb will construct a cell file by combining
all those tokens which are identically coded on the independent variables. This
file, Nepean.Cel, appears in Figure 6. The format of a cell file is discussed
in 6.
Figure
6.
At the same time as the cell file is being
constructed, certain information is written to the result file, Nepean.Res.
This includes the date and time, the name of the token file, the condition file
in its entirety and various statistics computed for Nepean.Tok, Nepean.Cnd and
Nepean.Cel. See Figure 7.
Figure
7.
Finally, choose Binomial, 1 level in the Cells menu. This will carry out the actual
variable rule analysis on Nepean.Cel. This and other options will be discussed
in 6. The results of the analysis are written to Nepean.Res, which already
contains some results from the cell construction procedure. See Figure 8. In
addition, GoldVarb draws a Scatter-gram, shown in Figure 9, comparing
the proportion of plural markers actually expressed in each cell of Nepean.Cel
with the proportion predicted by the statistical model constructed by the
program.
In this section we have followed the pathway from token
data (contained in the token file) through cell file (constructed according to
definitions in the condition file), to the results of variable rule analysis
(contained in the result file). In the following sections we will examine each
of these types of files in more detail.
Figure
8.
Figure
9.
4. TOKEN FILES
In 3 we saw how to use a previously saved token
file. It may be opened by clicking on its icon (this starts up GoldVarb), or by
choosing Open from the File menu after GoldVarb has been started. We will use
the term native to refer to a
file created or modified by GoldVarb. Such a file will be represented in the
Finder by one of the four distinctive GoldVarb document icons (see Figure 3).
The user who begins working with GoldVarb without
having a native token file will need to proceed in one of two ways:
(a) Use
the New command to create a new
empty document into which tokens will be typed, or
(b) Use
the Import command to import a
non-native token file, i.e. a file created by another application.
These two approaches are discussed in 4.1 and 4.2.
In version 2 of GoldVarb, a new feature, discussed in 4.3, allows the creation
of a new token file from existing token and condition files. Finally in 4.4 we
discuss another new feature introduced in version 2: searching and replacing
factors in the token file.
4.1 Creating a New Token File
When New
is chosen from the File menu, the
user is first given the opportunity to name the document which then appears on
the screen with the Factor specification box below it. The document will be empty except for a single open
parenthesis, prompting the user to type a token. Here are some guidelines for
the entering of tokens:
Each
token must be contained on a single line and must begin with an open
parenthesis (which is not considered to be part of the token).
The
end of a token is encountered the first time one of the following occurs:
* the
number of characters read is equal to the number of groups specified in the Factor
specification box, or
* the
end of the line is encountered, or
* a
space is encountered. (Thus the token may not contain any spaces, but comments may be placed at
the end of the line.)
The
token may contain the character . which will be replaced by the default factor
for the appropriate group when the tokens are checked.
If
the number of characters in the token is less than the specified number of
factor groups, the token will be filled with the appropriate fill character (as
specified by the Set fill character menu item discussed in 8.4)
Any
line not beginning with an open parenthesis is considered to be a comment and
is ignored.
When
a carriage return is typed, an open parenthesis is automatically inserted at
the beginning of the new line, thus preparing for the typing of a new token. If
one wishes to type a comment instead, simply backspace over the parenthesis.
Figure
10.
In addition to the tokens, the user must enter the
factor specifications in the appropriate box.[7] These are a set of declarations indicating which
factors may appear in each group (i.e. each column of the token file) as well
as the total number of groups. Here are some guidelines for the entering of
factor specifications:
The
Factor specification dialogue box
(Figure 10), must be the active window (i.e. its title bar must be filled).
The
button Set N of Groups allows
the user to enter or change the number of factor groups.[8]
For
each group, enter in the rectangular slots the legal factors for that group and
the default value which will be used to replace the character .. These
characters are shown in the dialogue box using the same font as in the token
document.
The
character / means does not apply.
The
TAB key may be used to move between the two rectangular slots.
Use
the small vertical arrows to move from group to group.
If
one changes the default or legal values, then clicking the button Modify
group will make the change
permanent.
The
default button Check all will
check all the groups, making sure there are at least two factors per group and
that the default values are legal. This is also done automatically when the
menu command Check tokens is
executed.
After both tokens and factor specifications have been
entered, the user should choose Check tokens from the Tokens menu. This command (which is executed automatically when cells are
created from tokens and conditions) determines whether all tokens contain only
legal factors, extends any short token with the appropriate fill character, and
replaces the character . by the appropriate default value.
4.2 Importing a Non-Native Token File
If the user has already prepared a token file using a
program other than GoldVarb, this file may be processed by starting up GoldVarb
and using the Import command. This
command, unlike Open, allows the
user to open any text file
created by any application. Assuming that the format of this file conforms to
the guidelines listed in 4.1, there is a short-cut which avoids the sometimes
laborious task of entering the factors in the Factor specification box. When the user chooses Generate factor
specs from the Tokens menu and clicks Yes in the ensuing dialogue box, GoldVarb scans the
tokens and builds up a list of factor specifications based on the occurrences
of factors in the token file. We can think of this command as a sort of inverse
of the command Check tokens
discussed above. That is:
Assuming
that the factor specifications are correct, Check tokens determines whether the tokens conform to them, i.e.
it makes sure that the tokens contain no illegal factors.
Assuming
that the tokens are correct, Generate factor specs determines factor specifications which conform to
them, i.e. it builds up lists of legal factors.
It is far easier to check the factor specifications
by eye than to enter them or, of course, to check the tokens by eye.
If an imported file is saved by GoldVarb, it becomes
native, i.e. it adopts the appropriate GoldVarb document icon as if it had
been created by the program.
4.3 Creating a Token File from an Existing Token
File
Normally one uses a token file, subject to recodings
defined in a condition file, in order to create a cell file. However, a new
feature of GoldVarb version 2 allows one to recode a token file in order to
create not a cell file but a new token file. This is convenient if one wishes
to do several recodings of the same data using complicated sets of conditions
which differ only slightly. A single initial recoding can be used to create a
new set of data (tokens) to which much simpler conditions will subsequently be
applied.
To use this feature, choose Recode to new token
file from the Tokens menu. When the
recoding has been completed, the old token and condition files will be closed
and the newly created token file will be opened on screen.
4.4 Searching and Replacing
For the purposes of preparing and correcting the raw
data, i.e. the tokens, GoldVarb provides several search and replace functions.
These are accessible via the last two items in the Tokens menu. When the item Search
& Replace is chosen, the
dialogue box shown in Figure 11 will appear on screen. As the title of this
little window implies, the search and replace functions apply only to the token
document, and searching is columnar, i.e. GoldVarb will search for a factor or
factors starting only in a particular column. This dialogue contains four
buttons, a pair of arrows used to change the column number, and two slots for
entering the string to be sought and its replacement if any. If, for example,
one wishes to search for the factor s in the third group, then type s in the
first rectangular slot, and then use the little arrows in the bottom right
corner to set the column number to 3. To begin searching, click the button Find, or use the command key 
`. The information about what is being sought will
also appear in the last item in the Tokens menu illustrated in Figure 24.
The search example just described is very simple,
involving only a single character. A less trivial example would be to search
for, say, the string nc starting in column #2 — that is, to search
for the simultaneous occurrence of the factor n in the second group and the
factor c in the third group.
Here are some guidelines for searching:
The
program searches only for strings beginning in the specified column, and only
in the token document.
The
open parenthesis is not
considered part of the token — for example, column #1 means group #1,
i.e. the first character after the parenthesis. Comments (lines not beginning with an open parenthesis) are ignored.
The
search begins on the line just after the start of the current selection and
continues to the end of the tokens or until the desired string is found. Thus
it is important to place the caret at the beginning of the tokens if the the
entire document is to be searched.
The
length of the string sought may not exceed the maximum number of groups allowed
in a token (30 in the current version of GoldVarb). The string must be
contained on a line, i.e. it may not contain any carriage return.
The
search is literal only — for example, the characters a, A, & are all
distinct. A scan for one of them will not find any of the others.
Let us suppose that the user wishes to replace all
occurrences of the factor s in group #3 by . To do so, they would type
this latter character in the second rectangular slot in Figure 11 and then hit
the button Change all. The Replace
All dialogue box of Figure 12 will
then appear, asking the user to confirm or cancel the requested operation. If
the user chooses Continue, then
GoldVarb will proceed to replace all such occurrences, and the dialogue box
(called the monitor) in Figure 12
will remain open until the operation is completed. (If the document is very
long, it is possible to switch to another application under the MultiFinder and
let GoldVarb continue in the background. The user will be informed when
GoldVarb has completed its task.)
Figure
11
Figure
12
The buttons Change and Change & Find in Figure 11 are used for replacing one occurrence
at a time. The former is undoable, using Undo in the Edit menu, as if the replacement had been
typed from the keyboard. The latter is not undoable and is equivalent to performing Change followed by Find.
Note that after performing Change all, only the last modification may be undone using the Undo command.
When using any of these four functions, the text in
the token document scrolls automatically so that each occurrence becomes
visible as it is found.
5. CONDITION FILES
A condition file is necessary when generating cells
from a token file. It specifies which groups to use as dependent and
independent variables, and how to recode the tokens. By specifying factor
groups and recodes within the condition file, the user is able to select and modify
the coding strings used to build cells without changing the data in the token
file. Since only those factor groups specified in the conditions are used to
build cells, some factor groups within the coding string can be used, for
example, as flags for sorting the data rather than as variables for variable
rule. The recodes allow the user to select the factor groups to be used both
for the dependent variable and for the independent variables, to select the
tokens to be used to build cells, to combine factors within a factor group, to
create new factors within a factor group, and to create new factor groups.
As with token files, a condition file may be created
by GoldVarb using New, or it may
be created by another program and imported into GoldVarb using Import. Many users find it difficult to master the syntax
required for the construction of a condition file. Thus GoldVarb includes a
simplified procedure for this task — yet another way of creating a
condition file — implemented through the Recode setup command in the Tokens menu. There is also the No
recode option discussed in 3.
5.1 Using the Recode setup dialogue box
We illustrate some simple recodings on the Nepean
plural data. Open Nepean.Tok and choose the Recode setup command in the Tokens menu. After asking the user to
enter a name for the condition file which GoldVarb will generate (we suggest a
& n vs. d for reasons which will soon be evident) the recoding dialogue
box in Figure 13 appears.
On the left side of this box we see a list of the
groups and factors in the token file. A list of recoded groups will be build up
on the right side. The six buttons in the middle (discussed in Table 1 below)
perform various forms of recoding; they are all initially deactivated because
none of the groups has as yet been selected. GoldVarb uses the empty space at
the bottom of the dialogue box to display occasional messages telling the user
what to do next.
Suppose we have reason to ignore the distinction
between adjectives and nouns in their effect on plural expression, and to
concentrate on the difference between the determiners and the other two
categories. Click #1 in the column on the left of the left hand list, and
then click Copy. The group with factors 1 and 0 then
appears in the right hand list. Now click #2, followed by Recode. The
group with factors a, n and d appears in the right hand list, with the a flashing.
Type a. Now the n starts flashing. Type a again to indicate that the
nouns are being reclassified with the adjectives. The screen should look like
Figure 14, with the d flashing. Type d. Finally, copy the #3 factor group to the right
hand list and click OK. GoldVarb automatically generates a new condition
file, shown in Figure 15.
Figure
13.
Figure
14.
Figure
15.
Let us consider another example using the Recode
setup command. We begin in the same
way as above, assigning a file name (here we suggest Nouns with subjects) and
then copying group #1 from the left to the right side. But this time let
us construct a new factor group by combining groups #2 and #3 into two
new factors: x will represent tokens which consist of a noun in
subject position (i.e. those tokens with n in column 2 and s in column 3) and y will
represent all other tokens. To do this, select groups #2 and #3 on the
left, then click in the AND button to indicate that we are interested in
simultaneity of factors in these two groups. Still on the left side of the box,
select the factors n and s and type x which will appear as the first factor in group #2 on
the right. Since this is the only
combination required, click in the space at the bottom as instructed.
The dialogue box should now appear as in Figure 16
with the small black rectangle beside the x flashing, indicating that we must type one more
letter as the recode value for all other combinations, that is, all tokens
which do not consist of a noun in
subject position. Type y, and we are done. Click in the OK button.
The dialogue box disappears and the newly generated condition file (Figure 17)
appears on screen.
A word about the numbering of groups: In Figures 14
and 16 the groups on the right side of the dialogue box (Groups after
recoding) are numbered consecutively starting at 1. The number appearing in
square brackets indicates the groups origin, i.e. the groups number before
recoding. Thus the notation 1[1] means that recoded group number one was
taken from group #1 on the left side. If the letter n appears instead of a
number, this indicates that the group is a new one which did not exist as such
in the token file. Thus the notation 2[n] in Figure 16 indicates that the
second recoded group was constructed from a combination of several groups (in
this case, two) on the left. In the resulting condition files (Figures 15 and
17), the original group number is shown at the beginning of each set of recode
conditions. The number 0 is used for new groups, as shown in Figure 17.
Figure
16.
Figure
17.
Finally, it should be noted that the Recode setup dialogue allows easy construction of only the
simplest and most common forms of recode. More complicated recodes must be
entered by typing directly into the condition file window. Nevertheless, the
two methods may be combined — i.e. one may generate a preliminary
condition file using the dialogue and then modify or extend it by typing.
See Table 1 for an
explanation of the six buttons in this dialogue box.
>> Copy >> Copy
the selected group(s) from the left to the right.
Exclude Exclude the factor(s) selected on the left. The
condition file will contain recoding conditions which will exclude from the
cells all tokens which contain an excluded factor.
>> Recode >> Recode
the group selected on the left and insert the resulting group on the right.
>> AND >> Combine
the two or more groups selected on the left into a single new group on the
right, using the predicate AND.
>> OR >> Combine
the two or more groups selected on the left into a single new group on the
right, using the predicate OR.
Remove << Remove
the group(s) which is/are selected on the right. This is useful for correcting
errors.
Table
1: Buttons in the Recode setup dialogue box
NOTE:
On
each button, the arrows, if any, indicate the direction of the operation which
the button performs.
The
list on the left side of the Recode setup box shows the groups contained in the token file, as specified in the Factor
specification box. The list on the
right side contains the recoded groups, i.e. the groups which will be generated
by the condition file which this dialogue box builds.
A
group or factor is selected or de-selected by clicking with the mouse. A factor
is selected if it is shown on a black background, while a group is selected if
its group number is shown on a black background. It is possible, for example,
to select a group without selecting any of its factors, but the opposite is not
possible. At most one factor at a time may be selected in a selected group.
A
button is activated only if an appropriate group, set of groups, factor, or set
of factors has been selected. For example, the Copy button is activated when
one or more groups are selected on the left, while Remove is
activated only if one or more groups are selected on the right.
5.2 The Application Values
Along with the condition file, GoldVarb must be told
what values of the dependent variable (the first recoded group), are pertinent
for constructing cells. During the final stages of cell creation, i.e. just
after processing the token and condition files, GoldVarb will display a
dialogue box in which the user is asked to enter the application values. If
only one value is entered, then this will be the application value, and all
other factors in the group will be counted as non-applications. If more than
one value is specified, only these values will be used, while recoded tokens
with any other factor in column one will be ignored. The maximum number of
values of the dependent variable is 9.
Example:
Suppose the dependent variable has factors abcd (after recoding). Consider
the following choices:
a: binomial
case; a
= application; b, c & d = non-application.
ab: binomial
case; a
= application; b = non-application; c & d omitted.
abc: trinomial
case; d
omitted.
If the user wishes to use any other combination of
factors (e.g. a & b applications, c & d non-applications), this must be
done by first recoding.
Variable rule computations are possible for only the
binomial case (in the current version of GoldVarb). However cell file creation
and cross-tabulation are possible in all cases. The adjustment of condition
files to eliminate knockout and singleton factors is discussed at the end of
6. Condition file syntax is discussed in further detail in the Appendix.
6. CELL FILES
A native cell file is created by GoldVarb from a
token file and a condition file using the Load cells to memory command in the Cells menu. A previously saved cell file may be opened by
double-clicking on its icon (this starts up GoldVarb) or by choosing Open in the File menu after starting
GoldVarb. However, as with token and condition files, the user can create a new
cell file by choosing New and
then typing directly into the new window, or a non-native cell file created
elsewhere may be imported using the Import command. In order to facilitate data importation,
the format of GoldVarb cell files has been chosen to be compatible with the
format used by Susan Pintzuks IBM-PC programs.
A cell file, such as the one depicted in Figure 6,
consists of the following parts:
1) On
the first line, the number of variants (up to 9) of the dependent variable in
column 1, followed immediately by a list of these variants.
2) On
the second line, the number of factor groups, not including the dependent
variable, right-justified in columns 1 and 2.
3) One
line for each of these factor groups, the number of factors in the group
right-justified in columns 1 to 4, immediately followed by a list of the
factors in the group.
4) In
the subsequent lines, the cells, each cell occupying two lines.
a) in
the first line, the number of tokens, right-justified in columns 1-4, 5-8,
9-12, etc., for each variant of the dependent variable. In the binomial case
there are two such data: the number of applications and the number of
non-applications.
b) in
the second line, the cell factors, one for each factor group.
5) The
end of the cell list is indicated by the value -1 in columns 3 and 4.
When GoldVarb creates a cell file from tokens and
conditions, the date and time and the names of the token and condition files
are inserted at the end of the list of cells, i.e. after the line containing
the -1.
With some data sets a knockout factor or a singleton may be flagged by the program beside the tabular
results (cf. Figure 7) which show the counts of factors in the cells. A
knockout is a factor for which applications occur with frequency 0% or 100%. A
singleton is a group which contains only one factor. Variable rule computations
cannot logically be performed on a cell file which contains knockouts or
singletons. One must generate a new cell file by recoding the original tokens
using a different condition file.
7. VARIABLE RULE ANALYSIS
Variable rule analysis can be performed provided
that:
(1) cells
have been successfully loaded into memory (either generated from tokens and
conditions or read from a previously created cell file), with neither knockouts
nor singletons, and
(2) only
one or two values have been declared in the Choose application value(s) dialogue box. This is the binomial case.[9] When only one application value is declared, this is
still binomial since the factors in the first recoded group are split into two
sets: the factor which is the application value, and all other factors, which
count as non-applications.
In the Cells menu, the items which perform variable rule analysis are: Binomial,
Up & Down, and Binomial, 1
level. The latter performs an
analysis based on all groups and all cells. An example of this was discussed in
3 and illustrated in Figures 8 and 9.
In Figure 8, the line starting with Iterations
keeps track of GoldVarbs progress in finding the maximum likelihood
estimation of the factor weights to a certain degree of accuracy, at which
point convergence is indicated. If the number of iterations reaches 20
without convergence, no further iteration is attempted and the current values
of the estimates are presented.[10]
A new feature of GoldVarb is the option of comparing
the log likelihood of a run and the maximum possible value of such a
likelihood. The usefulness of this test remains to be evaluated.
The scattergram (Figure 9) drawn at the completion of
a 1-level analysis may be printed, or it may be copied to the clipboard and
subsequently pasted into the Scrapbook or into a document of a graphics
application in order to save it. Further, while the scattergram window remains
open on screen, detailed information about any data point in the scattergram
may be obtained (and optionally written to the result document) by positioning
the cross-hair cursor over the point and clicking with the mouse button. The
size of each point is proportional to the number of tokens in the corresponding
cell(s), so that a large point far from the diagonal suggests interaction among
its factors.
For compatibility with some previous variable rule
programs, GoldVarb also displays, at the end of a 1-level analysis, the
Chi-square contribution from each cell as well as the average Chi-square per
cell.
Binomial, Up & Down performs a step-by-step analysis, at each level of
which only a subset of the factor groups are included and cells are contracted by
combining together in one cell all those which differ only in excluded groups.
At level 0 no groups are included so the cell list contracts into a single
cell, at level 1 only one group is included, at level 2 two groups, etc.
GoldVarb begins at level 0 and steps up until it finds no group whose inclusion
would significantly (p < 0.05) increase the log likelihood. GoldVarb then
starts again but at level n (n = the number of independent groups) at which
all groups are included and all cells are used without contraction (as in the Binomial,
1 level analysis), and steps down to
lower levels until it can no longer find a group whose exclusion does not
significantly decrease the log likelihood.
The results from the Best stepping up run (see
Figure 19) are usually identical to those from the Best stepping down run.
When they are not identical, this indicates some uncertainty about the status
of the factor groups included in one analysis but excluded from the other.
Figure 18 shows the Macintosh screen during a Binomial,
Up & Down analysis. In addition
to the menu bar, the figure shows two windows. The large window is the result
document. The small window above it is a dialogue box, called the monitor, indicating the status of the computation and
including buttons which allow the user to cancel or temporarily to suspend the
analysis. The horizontal bar is filled in as the computation proceeds. In the
figure it is less than half filled, since the step-up part has not yet been
completed.
As the analysis in Figure 18 proceeds, the cursor
rotates, imitating a rolling beach ball. If the user hits the Pause button, the analysis will be suspended and the
cursor will take the form 
until the computation is either resumed
by hitting Continue or cancelled
by hitting Cancel.[11]
If the program is running under the MultiFinder, then
during a variable rule compu-tation (or while pausing), the user may switch to
another application by clicking in the little GoldVarb icon in the upper right
corner of the screen. This allows one to use the computer for other purposes
while the analysis continues in the background. The pause feature is useful in
order to allow another application near–exclusive use of the CPU on a
temporary basis. If GoldVarb completes its work in the background, the user
will be alerted. The user may switch back to GoldVarb by clicking in any of its
windows (e.g. the monitor or the result document), or by clicking (possibly
more than once) in the little icon in the top right corner of the screen, or by
using the � menu.
In addition to the three buttons just described, the
monitor in Figure 18 also contains a small control giving access to GoldVarbs Auto-save feature. This feature is equivalent to the option
Automatically save textual results which appears in the Editing options dialogue box accessible via the Edit menu discussed
in 8.3. It is included in the monitor because the menus are inaccessible
during variable rule analysis.
Figure 19 shows the result document as it appears
immediately after completion of this Up & Down analysis. Unlike the 1-level
case, a scattergram is not drawn
at the completion of an Up & Down analysis. If a scattergram is desired, a
recoding must be done so that only the appropriate groups are used to make
cells for a Binomial 1-level
analysis.
Figure
18.
Figure
19.
8. USING THE MENUS
In this chapter we
describe each of GoldVarbs eight menus.
8.1 The � Menu
In addition to desk accessories, this menu contains
the item About GoldVarb. which,
when chosen, displays a box containing information about the program and its
authors. Some general documentation summarizing this manual is available by
clicking the Help button in this
box. More specific information about the various GoldVarb windows is available
through the Info & Help command in the Window menu, discussed in 8.6.
8.2 The File Menu
The File menu, illustrated in Figure 20, contains
commands for opening, closing and printing files.
Figure
20
The commands New, Open,
Import, Save and Save as are for the three types of data files (tokens, conditions and cells)
and for result files. With Open
only files created by this program can be opened, while Import allows one to open any text file created by any
application. The length of documents is limited only by the amount of memory
(RAM) available. When the user closes a GoldVarb file which has been modified,
or when Save is chosen from the
File menu, the ensuing dialogue box displays the icon of the document so one
can tell at a glance what type of file is to be saved.
When a token file is opened, the Factor
specification dialogue box which
appears is for entering groups and factors which will be used to check the
tokens before recoding. On the
other hand, the groups and factors listed at the beginning of a cell file are
only the independent groups obtained after recoding. For information about the format for entering data, see the Info & Help command in the Window menu.
The Close
command applies to the active (topmost) window, which may be one of the three
data types mentioned above, or some other document such as the clipboard, the
windows for displaying results (there are two: one for textual results, the
other for pictorial results), the window which displays documentation, or the
dialogue box for searching.
Only text results, not pictorial, can be saved to a
disk file with this version of GoldVarb. However, a picture — such as a
scattergram — may be printed. It may also be copied to the clipboard and
then pasted into the Scrapbook, or into a document in another application such
as MacPaint or MacWrite.
The command Print setup displays a standard dialogue box which allows the
user to select some basic options, such as the page orientation, for the
printer which is currently chosen.
GoldVarb has two main printing functions. The command
Print selection (shown chosen in
Figure 20) is used to print the selection in a text document. If no text is
selected, then this function is disabled. The next command Print document is used to print an entire text or pictorial
document. Either item will cause the dialogue box of Figure 21 to be displayed.
In this figure the items in the dotted rectangle are proper to GoldVarb while
the rest depend on the type of printer currently in use. The button Page
layout gives access to another
dialogue box (not illustrated here) which allows the user to change the margins
and choose a header and/or a footer. If the option Preview printing,
page-by-page is chosen, a representation of each printed page will appear on
the screen so it can be viewed before being sent to the printer.
A picture will always be printed on a single page,
with vertical or horizontal reduction if necessary. When printing text, the
number of lines per page depends on the font, font size and font style used to
display the document, as well as on the choices of margin, header and footer.
The total number of pages of text to be printed is not precalculated in the
current version of GoldVarb (hence the question mark in Figure 21) but will be
in a future version.
Figure
21
The command Transfer, an alternative to Quit, allows one to go directly to another program
without going first to the Finder. For example one may wish to transfer to
another application in order to process result files created by GoldVarb or to
prepare data files (although this can also be done within GoldVarb). Of course,
if the Macintosh is operating under the MultiFinder then other programs can run
simultaneously with GoldVarb, in which case there is no need to transfer or quit
in order to switch applications.
8.3 The Edit Menu
Figure
22
GoldVarb allows text editing in the token, condition
and cell windows, and optionally in the result document. A single click with
the mouse button sets the insertion point for typing. Double-clicking with the
mouse button will select a word, while triple-clicking will select a line. If
line numbers are displayed in a window in which text-editing is enabled, then
several lines may be selected by dragging the cross-cursor over the appropriate
line numbers while holding down the mouse button. The clipboard displays either
the last piece of text copied (which will be inserted if the Paste command is executed) or the last picture copied, for
example a scattergram or a cross-tabulation.
The Edit menu implements the standard text-editing
commands, plus a few options. The command Undo (which changes to Redo when appropriate) allows one to undo or redo the
last Cut, Copy, Paste
or Clear operation, or the last
sequence of characters typed from the keyboard, including backspaces. However,
if the user leaves a window in order to work in another, and then return to the
first, it will then be impossible to undo/redo Cut, Copy
or Paste, because the contents of
the clipboard may have been changed. However, keyboard entries and Clear remain undoable/redoable.
The appearance of the Undo item in this menu changes, depending on what
operation, if any, can be undone or redone. In Figure 22 it reads Undo Copy, indicating that the user has just copied some text
to the clipboard. If this copy is undone, then the text previously stored on
the clipboard, if any, will be restored. This restoration of the old clipboard
also occurs when Cut is undone.
However, if the previous contents were a picture, they cannot be restored by undoing a Cut or Copy
operation.
The item Line numbers allows the display of line numbers in the left
margin of any window containing text. It works like a toggle switch. A check
mark appears on the left if it is on.
Finally, the last item in the Edit menu, Editing
options, causes GoldVarb to display
the dialogue box illustrated in Figure 23. This dialogue allows the user to set
several text-editing options.
Automatically
save textual results, when checked,
causes the text in the result document to be automatically written to the
appropriate disk file during variable rule analysis. This is useful protection
against unexpected interruptions, such as a power failure or a fatal error
occurring in another program running simultaneously under the MultiFinder. This
option is also available in the monitor as discussed in 7.
Allow
editing of textual results, when
checked, enables text-editing in the window used for text results. By default
this option is chosen, but it may be turned off in order to prevent inadvertent
modification of the result document. (No editing is possible in the pictorial
result window used for cross-tabulation and scattergrams.)
Automatic
indentation, when checked, means
that when a carriage return is typed the new line will be indented with the
same number of spaces (if any) as in the line just above it. If the Option
key is help down while the return key is pressed, automatic indentation is
temporarily suppressed if currently active, or temporarily activated if not.
Replace
tab by blanks, when chosen, means
that when a tab character is typed in a document, the specified number of
blanks is inserted instead of the tab character (ASCII code #9). This has no
effect on any tabs already present in the document.
Figure
23
8.4 The Tokens Menu
This menu, enabled only if a token document is open,
is illustrated in Figure 24. It contains a variety of commands which allow the
user to verify, recode or modify the tokens.
Generate
factor specs is used to scan
through the tokens in order to determine what factors they contain. These
factors will be displayed in the Factor specification dialogue box, replacing any factors currently
stored. This is convenient when importing an already existing token file, so
that one does not have to type the factor specifications directly into the box.
Show
factor specs causes the groups and
factors (which were entered one-by-one in the dialogue box, or which were generated using the above command) to be
displayed in their entirety in the result window.
Set
fill character allows one to
specify the character to use to fill short tokens.
Check
tokens verifies that each column in
each token contains a valid factor, replaces the character . by the default
character for the appropriate group, and fills short tokens with the specified
character.
Note the difference between the commands Generate
factor specs and Check tokens described above. The former assumes that the tokens
are correct and uses them to determine the number of groups and what factors
are used for each group, while the latter command uses the factor
specifications in the box in order to determine whether the tokens conform to
these specifications.
No
recode generates a condition file
which corresponds to inclusion of all groups and factors from the token file
when creating cells.
Recode
setup is a short cut for
constructing a condition file, allowing one to define in a dialogue box several
basic kinds of recode without having to learn the syntax for entering
conditions. It is discussed in 5.1.
Recode
to new token file allows the user
to create a new token file from an existing token file and a condition file.
The old token and condition files will be closed and the newly created token
file will be opened on screen. This command is an alternative to the command Load
cells to memory in the Cells menu
discussed in 8.5. The latter is used to create a cell file from tokens and
conditions.
Note that each of these three commands just described
creates a new GoldVarb file. In the menu, the icon which accompanies the
command indicates the type of file which will be created. The same is true of Load
cells to memory in the Cells menu.
Finally, the last two items in the Tokens menu deal
with searching and replacing as discussed in 4.4.
Search
& Replace opens the dialogue
box illustrated in Figure 11.
The
appearance of the last item changes, depending on what character string, if
any, has been entered in the search-and-replace dialogue box. If no such string
has been entered, then this item reads Find next, but is disabled. In Figure 24 this item reads Find
next s @ col. #3 — this is
how it would appear if the search-and-replace dialogue appears as in Figure 11.
Figure
24
8.5 The Cells Menu
The Cells menu performs various operations on cells,
including variable rule analysis. The menu is divided into four parts. The
first part contains four items:
Load
cells to memory is the most
important item since it is used either to generate a new cell file from tokens
and conditions, or to load cells from a previously created cell file. In either
case, counts of the groups and factors in the cells are written to the result
document in tabular form.
Once
cells have been loaded using the above command, Show cells in memory causes a list of them to be written to the result
document.
Show
application values puts up a small
dialogue box displaying the application values which are currently in use.
Cross
tabulation allows the user to
cross-tabulate the factors in any two groups. The table may be displayed either
as a picture in a separate window as in Figure 26, or as text which will be
written to the result document. The pictorial method gives a more attractive
table which, although it cannot be saved with the result document, can be
printed or exported (see 8.3).
Figure
25
The second set of items in the Cells menu performs
variable rule analysis on the cells currently in memory. Binomial, Up &
Down and Binomial, 1 level are discussed in 7. Multinomial variable rule
analysis (i.e. with more than two application values) has not yet been
implemented. Hence the item Multinomial, 1 level is not functional.
The third part of the Cells menu is used to set
various options. Each functions like a toggle switch, with a check mark
displayed on the left if the option is on.
Rapid
computation greatly accelerates
calculations by sacrificing some accuracy. Results obtained with and without
this option should nevertheless agree to about the third or fourth digit after
the decimal point.
Centre
factors determines which method will
be used to determine the average probability of the factors in a given group.
When this option is chosen, each factor in a group is given equal weight.
Otherwise each factor is weighted according to its occurrences relative to
total occurrences of all factors in the group. (By occurrences is meant the
sum of applications and non-applications.)
Show
model fit causes GoldVarb to display
the maximum likelihood and Chi-square fit during variable rule computations
involving more than one independent variable.
An
explanation of the option Shut up
is left as an exercise for the reader.
Finally, the last command in the Cells menu, Show
memory info., is of interest only to
the technically oriented user. It displays in the result document various
information about GoldVarbs use of dynamically allocated memory.
Figure
26
8.6 The Window Menu
This menu contains a variety of commands controlling
the arrangement of, and giving information about, the windows on the screen.
If
two or more windows are open, the Change window command selects the bottom window, making it active.
This is handy for uncovering a window obscured by others.
Info
& Help displays information
about the active window. For example, if the cell window is topmost, then
information about the format of cell data will be displayed, as well as the
number of lines and characters contained in this window.
Zoom expands the active window to fill the entire screen,
or shrinks it down to its former size.
Tile
repositions all windows so that the
entire screen area is used but no overlap of windows occurs.
Restore
positions causes all windows to be
restored to their default positions.
In
a text document, Home cursor
positions the caret at the beginning of the text, while Bottom cursor positions it at the end. These two commands have a
similar scrolling action on a window containing a picture.
Figure
27
8.7 The Font & Style Menus
The Font menu is used to select the font for the
active window. Non-proportional fonts such as Courier or Monaco are recommended
for tables so that their columns will be properly aligned. However, if the user
does a Cross tabulation (Cells
menu) and chooses Picture, the columns will be properly aligned regardless of
which font is used. For such pictorial results, the font or font size may be
changed only by recreating the windows contents; i.e. one must re-execute the
appropriate command after choosing a different font or font size.
Using the Style menu, illustrated in Figure 28, one
may change the size of the font used to display the text in the active window.
The appearance will be best if a size displayed in bold/outline in the menu is
chosen. For example, in the figure sizes 9, 10, 12, 14 and 18 will appear best.
Font sizes greater than 12 are not recommended for scattergrams.
The Style menu also contains several items which
control the style of the text. As with fonts and font sizes, the choice of
style applies to the entire
document, not just the selection. The style Bold may, for example, be used to improve readability
when displaying the Macintosh screen on an overhead projector. The style
Condense may facilitate printing a text document containing long lines.
Figure 28
9. GLOSSARY
active window The
topmost window. Its title bar is filled in. In an inactive window, the title
bar is white except for the title. An inactive window may be made active by
clicking anywhere in it. If the active window is a document (as opposed to a
dialogue box) then GoldVarb displays in its bottom left corner the amount of
memory (RAM) still available to the program. This information is also displayed
in the topmost document when it is not active, i.e. when covered by one or more
dialogue boxes.
apple menu The
first menu on the left. The first menu item is About GoldVarb which displays information about the authors and the
current version of the program, and allows access to some documentation. This
menu also contains desk accessories.
application 1. A computer program.
2. Presence of the appropriate value of the
dependent variable, as in application value.
caret The
tiny vertical bar which flashes indicating the insertion point in a text
document in which editing is enabled. It appears only when the selection is of
length zero. It is sometimes called the cursor, at the risk of confusing it with the mouse point.
close box The
small square in the upper left corner of document windows and some dialogue
boxes. Clicking in this box will close the window.
cursor The
small symbol which follows the movements of the mouse and which changes form
depending on what active window, if any, is beneath it. Also called the mouse
point. When over the contents of a
text document in which editing is enabled, it takes the form of an I-beam and
allows the user to change the position of the caret by clicking. When over a scattergram, or the line
numbers in a text document, it takes the form of a cross-hair. When over a
scroll bar, or when no active window is beneath it, it takes the form of an
arrow. During long computations, the cursor takes the form of a rolling beach
ball. When a long computation is temporarily suspended, the cursor takes the
following form: 
.
deactivated button In
a dialogue box, a button whose title is drawn in gray (rather than black) and
which is not currently functional.
default button In
a dialogue box, the button which is boldly outlined. Pressing the Return key or
the Enter key on the keyboard is equivalent to clicking with the mouse in this
button.
desk accessory A
small program which can be called using the apple menu and which is available
for use without leaving the current application. Examples: Calculator, Alarm Clock, Key Caps.
Finder An
essential component of Macintosh system software, the Finder is the user
interface of the Macintosh file system. It is a program which pictorially
displays the files on the disk(s) currently in use, with different icons for
different types of files, and which provides basic file management operations,
for example: placing files in folders (directories), copying files, deleting
files, copying disks, launching applications, etc.
grow box The
small symbol in the lower right corner of a document window. The windows size
may be changed by placing the arrow cursor in this box and holding the mouse
button down while dragging.
monitor The
dialogue box which is displayed during some time-consuming processes (replace
all, variable rule analysis) and which displays the status of the process. Its
title indicates the type of process. It contains three buttons — Continue, Pause
and Cancel — which can be
used to suspend or cancel the operation. As the operation proceeds (or while
pausing), the user may move or resize windows in GoldVarb. If the program is
running under the MultiFinder, the user may switch to another application.
mouse point See
cursor.
MultiFinder A
component of Macintosh system software which allows the user to run several
different applications simultaneously. For example, GoldVarb, the Finder and a
word processing program can be used simultaneously provided that the computer
has sufficient memory (RAM). By default GoldVarb requests 512 kilobytes of RAM
when running under the MultiFinder, although the user may easily change this by
entering the desired amount in the lower right corner of the Info window. To
open this window, select the GoldVarb application icon and choose Get Info from the Finders File menu.
native file A
file created or saved by the application. GoldVarb has four kinds of native
files, each with its own distinctive icon: token files, condition files, cell
files and result files. See Figure 3.
selection In
a text document or in a list in a dialogue box, the selection is shown on a
black background (or possibly on a gray or a colour background on a Macintosh
II). Selecting is done by clicking and dragging with the mouse. Text editing
operations such as Cut, Copy, Paste
and Clear apply to the currently
selected text; if no text is selected, then the insertion point is indicated by
the flashing caret. These comments apply to the active window only. When a GoldVarb text document is inactive, the
selection, if any, is outlined; if there is no selection then the caret is
invisible. When an inactive text document containing a selection is made
active, the outlined region is filled with the appropriate background colour.
title bar In
a document window and in some dialogue boxes, the bar across the top of the
window which contains the title. The windows position on the screen may be
changed by placing the arrow cursor in this bar and holding the mouse button
down while dragging.
APPENDIX I. More on
Condition Files
The following discussion of condition file syntax has
been adapted from Susan Pintzuks documentation of her IBM-PC variable rule
programs.
The data within a condition file is in the form of a
LISP list. Each element of the list is itself a LISP list consisting of two
parts: a group number (column number within the coding string) and an optional
set of recode conditions. If no recode conditions are specified, the group is
used exactly as it is coded in the token. All groups specified in the condition
file, and only those groups, are used to build cells. The first group in the
condition file list is used as the group containing the dependent variable. The
order of groups specified within the condition file determines the order of
groups within the cell.
Each recode condition is again a LISP list consisting
of two parts: the first part is the recode value, the value to be used for the
group for those tokens which meet the second part of the condition, the test
clause. The recode value is either a single character or NIL. If it is NIL
then tokens which fulfill the test clause are excluded when building cells.
Similarly, if the recode value for the dependent variable group is /, tokens
which fulfill the test clause are also excluded. If the recode value for an
independent group is /, the remaining groups for tokens which fulfill the
test clause are used in the construction of cells.
There are five test clause predicates: AND, OR,
NOT, COL and ELSEWHERE. Case is irrelevant for predicates, e.g. OR and
or are equivalent. However, case is significant for factors! For example b
and B are two distinct factors.
AND, OR and NOT are the standard logical
operators; AND and OR take two to 20 predicates as arguments, NOT takes a
single predicate as argument. If it is necessary to define a recode condition
with more than 20 arguments for AND or OR, two or more of the arguments can
be nested more deeply. For example:
(AND a1 a2 a3 a4 a5 a6 a7 a8 a9 a10
a11 a12 a13 a14 a15 a16
a17 a18 a19 (AND a20 a21 a22 a23 a24)).
COL takes two arguments, a group number (i.e. column
number within the coding strings) and a single character representing a legal
factor value for that group; COL is true if and only if that column of the
coding string contains the specified value.
ELSEWHERE is always true; it is used as the last test
clause within a set of test clauses for a group, and forces the recoding of the
group to the specified value if none of the previous conditions for that group
has been met.,
Here is an example of a condition file:
(
(4 (d (OR (COL 4 d) (COL 4 c)))
(s
(ELSEWHERE)))
(5)
(3 (/ (OR (COL 3 s) (COL 3 t) (COL 3 u)))
(m (OR
(OR (COL 3 n) (COL 3 h))
(OR (COL 3
1) (COL 3 2) (COL 3 3) (COL 3 w) (COL 3 u)
(COL 3 y) (COL 3 p) (COL 3 t) (COL 3 r) (COL 3 x))))
(x (AND
(OR (COL 3 n) (COL 3 h)) (COL 7 n)))
(NIL
(ELSEWHERE)))
; interactive group
(0 (1 (AND (COL 2 x) (COL 8 a)))
(2 (AND
(COL 2 x) (COL 8 b)))
(3 (AND
(COL 2 y) (COL 8 a)))
(4 (AND
(COL 2 y) (COL 8 b)))
(/
(ELSEWHERE)))
)
Lines with a semi-colon in column 1 are comment lines
and are ignored when processing the condition file. In the above example, group
#4 is the dependent variable while groups 5, 3, and 0 are the independent
variables which will be used to build cells. Group #5 has no recode conditions
and therefore will be used exactly as it is coded in the token file. Groups 2
and 8 are interactive factor groups; the interaction is investigated by
creating a new group, which is given the number 0.
Note that GoldVarb processes the recode conditions
for each group in the order in which they appear in the conditions. The first
condition that is satisfied for each token is used for recoding, and the rest
of the conditions for that group are ignored for that particular token. For
this reason, ELSEWHERE should be used only as the last condition for a group
-- it is always true, so any conditions listed after it (for the same group)
will be ignored. For example, if the conditions for group #5 are:
(5 (t (ELSEWHERE))
(a (COL 5 x))
(b (COL 5 w)))
then all tokens will be recoded t for group #5,
even if they were originally coded x or w.
Similarly, the user should make sure that a condition
with recode value NIL is placed correctly within the list of conditions.
Consider the following two sets of recode conditions for group #3:
1. (3 (a (COL 3 s)))
(b
(COL 3 w))
(NIL
(AND (COL 3 s) (COL 4 t))))
2. (3 (NIL (AND (COL 3 s) (COL 4 t)))
(b
(COL 3 w))
(a
(COL 3 s)))
In the first example, any token coded s for group
#3 will be recoded,a, and the remaining conditions will be ignored for that
token; therefore, the third condition can never be met. In the second example,
any token coded s for group #3 and t for group #4 will not be used to build
cells; therefore, only tokens coded s for group #3 and NOT coded t for
group #4 will be recoded a. If the user wishes to eliminate unconditionally
certain tokens from the cells, it is suggested that these recodes to NIL be
placed as the first recodes within the list for the dependent variable. For
example, if the dependent variable is factor group #2, and all tokens coded s
in group #3 and t in group #4 are not to be used to build cells, then the
first part of thecondition file might look as follows:
(
(2 (NIL (AND (COL 3 s) (COL 4 t)))
(a (OR (COL 2 x) (COL 2
y))) ...
When creating new groups which do not exist in the
original tokens (indicated by a group number of 0, as in one of the examples
above), the user should include an ELSEWHERE condition if there is a
possibility that for some of the tokens, none of the conditions for creating
the new group will be true; otherwise, the first such token will generate an
error message and cell creation will be aborted.
For those users not familiar with LISP syntax, note
that:
1) The
list of conditions, i.e. the entire contents of the file, must be enclosed
within a set of parentheses;
2) Each
element of the list, i.e. group number plus optional recode conditions, must be
enclosed in parentheses;
3) Each
recode, i.e. recode value plus condition, must be enclosed in parentheses;
4) Each
predicate must be enclosed in parentheses.
Other than the restrictions specified above, the
format of a condition file is fairly free: it is not necessary that any
particular element appear in any particular position on a line, since
parentheses completely determine the structure of the data within the condition
file. Individual elements within the file are terminated by a space, open parenthesis,
close parenthesis, or end of line. Comment lines may be placed anywhere in the
condition file. The close parenthesis for the entire condition file must be the
last character within the file, except for spaces, carriage returns, or comment
lines. All comment lines must be signalled by a semi-colon in column 1.
APPENDIX II. The
Macintosh character set
The Macintosh character set is an extension of the
familiar ASCII character set and includes a total of 256 characters, numbered
from 0 through 255. Of these, the first 32, numbered from 0 through 31, are
reserved for special control characters (for example: carriage return, tab,
form feed, etc.). This leaves a total of 224 characters available for use in
documents. The 52 upper and lower case letters of the English alphabet, as well
as basic punctuation characters, are contained in the range 32127. Characters
numbered 128 through 255 (sometimes referred to as high ASCII) are used for characters with diacritical marks and
for further punctuation as well as special symbols. In some fonts, many
characters — especially high ASCII characters — are not even
defined. Such characters appear on screen as nondescript little rectangles and
may be blank when printed.
The default font for GoldVarb documents is Courier
12. This font has a number of advantages: it prints well on laser printers, it
is monospaced (i.e. almost all its characters are of the same width), and it
contains very few undefined characters. This last property makes it especially
convenient for token files in which some groups have many different factors.
Characters 32 through 255 of the font Courier 12 are
illustrated in the table below. The ACSII code of any character in the table is
found by adding the small figure at the top of the column to the small figure
at the left end of the row. Note that three characters have special
significance in tokens: ( used to introduce a token, / indicating a token
or group to be excluded, and . which will be replaced by the default factor.
The space (#32) and the option-space (#202) should not be used in tokens. In
addition, the following ASCII codes correspond to undefined and/or invisible
characters: 127, 174, 190, 206, 207, 222, 223 and 228. With these exclusions,
more than 200 characters are still available. In the current version of
GoldVarb the maximum number of factors per group is 200.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
32 ! " # $ % & ' ( ) * + , - . /
48 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
64 @ A B C D E F G H I J K L M N O
80 P Q R S T U V W X Y Z [ \ ] ^ _
96 ` a b c d e f g h i j k l m n o
112 p q r s t u v w x y z { | } ~
128
144
160
176
192
208 – —
224
240 �
S
.